Re: Combination of ordered-set aggregate function terminates JDBC connection on PostgreSQL 9.6.5

Heikki Linnakangas <hlinnaka(at)iki(dot)fi> writes:
> On 10/12/2017 05:27 AM, Tom Lane wrote:
>> Speaking of AggGetAggref, there's another thing that I think 804163bc2
>> did wrong for ordered-set aggregates: it can return the wrong Aggref
>> when two aggregates' intermediate states have been merged.

> Conceivably we could have another function like AggGetAggref() that
> returns all of the Aggrefs. But I don't think it's worth the
> complication. If the transition function needs to do something different
> depending on the aggregate it's for, well, don't do that. Define a
> different transition function for both aggregates.

Thinking about it more clearly, if a transition function is being run
on behalf of several different Aggrefs (with identical input states),
then no, the transition function should not care which Aggref it
looks at. As you say it mustn't do anything different on the basis of
the finalfn-related fields. The problem occurs when a finalfn calls
AggGetAggref --- then, I think that the finalfn is entirely entitled
to expect that it will see its own Aggref, not some other one that
happens to share input+transition.

So the issue is that we need some different behavior during
finalize_aggregate than during the transition function calls.
I'm inclined to put back the curperagg field we had before,
and populate that during finalize_aggregate. Then, AggGetAggref
would look at either curperagg or curpertrans, whichever is set.

regards, tom lane